Fig. 1

Strength of purifying selection estimated among mammals versus gene connectivity in the human metabolic network. Nodes were divided using the 25th, 50th, and 75th percentiles and the mean ± standard error of the residuals of a linear regression of dN/dS controlling for genomic variables (CDS length, codon bias, and GC content) is plotted for each group. Global differences between groups were assessed by Kruskal-Wallis Rank Sum test. Highly connected genes are under stronger purifying selection